Techniques for detecting text

Abstract

In some examples, a system for detecting text in an image includes a memory device to store a text detection model trained using images of up-scaled text, and a processor configured to perform text detection on an image to generate original bounding boxes that identify potential text in the image. The processor is also configured to generate a secondary image that includes up-scaled portions of the image associated with bounding boxes below a threshold size, and perform text detection on the secondary image to generate secondary bounding boxes that identify potential text in the secondary image. The processor is also configured to compare the original bounding boxes with the secondary bounding boxes to identify original bounding boxes that are false positives, and generate an image file that includes the original bounding boxes, wherein those original bounding boxes that are identified as false positives are removed.

What is claimed is: 
     
       1. A system for detecting text in an image, comprising:
 a memory device to store a text detection model trained using images of up-scaled text; 
 a processor to: 
 perform, using the text detection model, text detection on an image to generate original bounding boxes that identify potential text in the image; 
 generate a secondary image comprising up-scaled portions of the image associated with bounding boxes below a threshold size; 
 perform, using the text detection model, text detection on the secondary image to generate secondary bounding boxes that identify potential text in the secondary image; 
 compare the original bounding boxes with the secondary bounding boxes to identify original bounding boxes that are false positives; and 
 generate an image file comprising the original bounding boxes, wherein those original bounding boxes that are identified as false positives are removed. 
 
     
     
       2. The system of  claim 1 , wherein the processor is to process the image file with a text recognition algorithm to generate a text document comprising machine encoded text. 
     
     
       3. The system of  claim 1 , wherein to compare the original bounding boxes with the secondary bounding boxes to identify original bounding boxes that are false positives comprises:
 determine whether a secondary bounding box has been generated for the portion of the image associated with a specific one of the original bounding boxes; and 
 if no secondary bounding box has been generated, identify the specific one of the original bounding boxes as a false positive. 
 
     
     
       4. The system of  claim 1 , wherein to compare the original bounding boxes with the secondary bounding boxes to identify original bounding boxes that are false positives comprises, for each of the secondary bounding boxes:
 compare the secondary bounding box with its corresponding original bounding box to determine a degree of similarity; 
 compare the degree of similarity with a similarity threshold; and 
 if the degree of similarity is below the similarity threshold, identify the corresponding original bounding box as a false positive. 
 
     
     
       5. The system of  claim 4 , wherein to determine the degree of similarity comprises to compute a Jaccard index for the secondary bounding box and its corresponding original bounding box. 
     
     
       6. The system of  claim 4 , wherein the similarity threshold is a Jaccard index of 0.8 to 0.9. 
     
     
       7. The system of  claim 1 , wherein the threshold size is a threshold height of less than 10 pixels, and the up-scaled portions of the image are up-scaled by a factor greater than 2. 
     
     
       8. The system of  claim 1 , wherein the memory device stores the images of the up-scaled text, and the processor trains the text detection model using the images of the up-scaled text. 
     
     
       9. The system of  claim 8 , wherein the images of the up-scaled text used to train the text detection model comprise text images with an original height less than 10 pixels that are up-scaled by a factor of 3 or more. 
     
     
       10. The system of  claim 1 , wherein the image is one of:
 a scanned document; 
 and an image captured by a camera. 
 
     
     
       11. A method of detecting text in an image, comprising:
 performing text detection on an image to generate original bounding boxes that identify potential text in the image; 
 generating a secondary image comprising up-scaled portions of the image associated with bounding boxes below a threshold size; 
 performing text detection on the secondary image to generate secondary bounding boxes that identify potential text in the secondary image; 
 comparing the original bounding boxes with the secondary bounding boxes to identify original bounding boxes that are false positives; and 
 generating an image file comprising the original bounding boxes, wherein those original bounding boxes that are identified as false positives are removed; and 
 processing the image file with a text recognition algorithm to generate a text document comprising machine encoded text. 
 
     
     
       12. The method of  claim 11 , wherein comparing the original bounding boxes with the secondary bounding boxes to identify original bounding boxes that are false positives comprises:
 determining whether a secondary bounding box has been generated for the portion of the image associated with a specific one of the original bounding boxes; and 
 if no secondary bounding box has been generated, identifying the specific one of the original bounding boxes as a false positive. 
 
     
     
       13. The method of  claim 11 , wherein comparing the original bounding boxes with the secondary bounding boxes to identify original bounding boxes that are false positives comprises, for each of the secondary bounding boxes:
 comparing the secondary bounding box with its corresponding original bounding box to determine a degree of similarity; 
 comparing the degree of similarity with a similarity threshold; and 
 if the degree of similarity is below the similarity threshold, identifying the corresponding original bounding box as a false positive. 
 
     
     
       14. The method of  claim 13 , wherein determining the degree of similarity comprises to computing a Jaccard index for the secondary bounding box and its corresponding original bounding box. 
     
     
       15. The method of  claim 11 , wherein the threshold size is a threshold height of less than 10 pixels, and the up-scaled portions of the image are up-scaled by a factor greater than 2. 
     
     
       16. A computer program product for detecting text in images comprising a computer readable storage medium having program instructions embodied therewith, wherein the computer readable storage medium is not a transitory signal per se, and wherein the program instructions are executable by a processor to cause the processor to:
 perform text detection on an image to generate original bounding boxes that identify potential text in the image; 
 generate a secondary image comprising up-scaled portions of the image associated with bounding boxes below a threshold size; 
 perform text detection on the secondary image to generate secondary bounding boxes that identify potential text in the secondary image; 
 compare the original bounding boxes with the secondary bounding boxes to identify original bounding boxes that are false positives; and 
 generate an image file comprising the original bounding boxes, wherein those original bounding boxes that are identified as false positives are removed. 
 
     
     
       17. The computer program product of  claim 16 , wherein to compare the original bounding boxes with the secondary bounding boxes to identify original bounding boxes that are false positives comprises:
 determine whether a secondary bounding box has been generated for the portion of the image associated with a specific one of the original bounding boxes; and 
 if no secondary bounding box has been generated, identify the specific one of the original bounding boxes as a false positive. 
 
     
     
       18. The computer program product of  claim 16 , wherein comparing the original bounding boxes with the secondary bounding boxes to identify original bounding boxes that are false positives comprises, for each of the secondary bounding boxes:
 compare the secondary bounding box with its corresponding original bounding box to determine a degree of similarity; 
 compare the degree of similarity with a similarity threshold; and 
 if the degree of similarity is below the similarity threshold, identify the corresponding original bounding box as a false positive. 
 
     
     
       19. The computer program product of  claim 18 , wherein to determine the degree of similarity comprises to compute a Jaccard index for the secondary bounding box and its corresponding original bounding box. 
     
     
       20. The computer program product of  claim 16 , wherein the threshold size is a threshold height of less than 10 pixels, and the up-scaled portions of the image are up-scaled by a factor greater than 2.